Skip to contentSkip to conversation

i386: fix singlestepping though a syscall

PATCH In review Chuck Ebbert 2 participants 2 messages Feb 17, 2006

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 i386: fix singlestepping though a syscall

Singlestep through a syscall using vsyscall-sysenter had two bugs:

1. Setting TIF_SINGLESTEP is not enough to force do_notify_resume() to be run on return to user; TIF_IRET must also be set.

2. do_notify_resume() was being passed masked copies of task_thread_flags, so TIF_SINGLESTEP was never seen as set when it was called. This was changed to use the 'test' instruction instead of 'and'; a duplicated piece of code was removed instead of fixing that part too.

Also changed some misleading 'jne' to 'jnz' to make it clearer what is happening.

---
 arch/i386/kernel/entry.S |   17 +++++------------
 arch/i386/kernel/traps.c |    1 +
 2 files changed, 6 insertions(+), 12 deletions(-)
arch/i386/kernel/entry.S +5 -12
@@ -152,9 +152,9 @@ ENTRY(resume_userspace)
 					# setting need_resched or sigpending
 					# between sampling and the iret
 	movl TI_flags(%ebp), %ecx
-	andl $_TIF_WORK_MASK, %ecx	# is there any work to be done on
+	testl $_TIF_WORK_MASK, %ecx	# is there any work to be done on
 					# int/exception return?
-	jne work_pending
+	jnz work_pending
 	jmp restore_all
 
 #ifdef CONFIG_PREEMPT
@@ -301,21 +301,14 @@ work_pending:
 	jz work_notifysig
 work_resched:
 	call schedule
-	cli				# make sure we don't miss an interrupt
-					# setting need_resched or sigpending
-					# between sampling and the iret
-	movl TI_flags(%ebp), %ecx
-	andl $_TIF_WORK_MASK, %ecx	# is there any work to be done other
-					# than syscall tracing?
-	jz restore_all
-	testb $_TIF_NEED_RESCHED, %cl
-	jnz work_resched
+	jmp resume_userspace
 
+	ALIGN
 work_notifysig:				# deal with pending signals and
 					# notify-resume requests
 	testl $VM_MASK, EFLAGS(%esp)
 	movl %esp, %eax
-	jne work_notifysig_v86		# returning to kernel-space or
+	jnz work_notifysig_v86		# returning to kernel-space or
 					# vm86-space
 	xorl %edx, %edx
 	call do_notify_resume
arch/i386/kernel/traps.c +1 -0
@@ -795,6 +795,7 @@ debug_vm86:
 
 clear_TF_reenable:
 	set_tsk_thread_flag(tsk, TIF_SINGLESTEP);
+	set_tsk_thread_flag(tsk, TIF_IRET);
 	regs->eflags &= ~TF_MASK;
 	return;
 }
Signature · 2 lines
Chuck
"Equations are the Devil's sentences."  --Stephen Colbert
#1 lore raw
Quoted 6 lines · Chuck Ebbert

Singlestep through a syscall using vsyscall-sysenter had two bugs:

1. Setting TIF_SINGLESTEP is not enough to force do_notify_resume() to be run on return to user; TIF_IRET must also be set.

Interesting, but I think you pinpointed the _real_ bug.

TIF_SINGLESTEP very much should be enough to force do_notify_resume() to be run on return to user space.

Sounds like somebody is testing _TIF_WORK_MASK rather than _TIF_ALLWORK_MASK.

I'd suspect the "work_pending" case. Looks like we miss testing the TIF flags there.

Oh, actually, I think you should just remove the clearing of _TIF_SINGLESTEP from _TIF_WORK_MASK. Does that fix the bug.

Linus